Antero-inferior glenohumeral instability is associated with an abnormal position of the coracoid process

Abstract

Background

The role of morphologic variation in the etiology of antero-inferior glenohumeral instability (GHAI) remains debated. Clinically, the coracoid process (CP) varies in size and position, and simple coracoid lowering (“Trillat procedure”) can prevent recurrent anterior dislocation. Additionally in some cases, fracture of the coracoid is necessary for an anterior dislocation to occur. Therefore, the CP position relative to the glenoid center was compared in stable and unstable shoulders, hypothesizing that the CP tip is more superior, more medial, and more anterior in patients with recurrent GHAI than in those with stable shoulders or posterior glenohumeral instability (GHPI).

Methods

Computed tomographic scans of shoulders of 60 patients with GHAI, age- and sex-matched to 30 shoulders of patients with GHPI and 13 unmatched individuals with healthy shoulders (control group) were semi-automatically segmented into three-dimensional models. All shoulders were scaled to uniform size, and left-sided cases were mirrored to the right side. A statistical shape model was created for each cohort, and anatomic landmarks were generated automatically. Models were aligned to the glenoid plane and center, and three-dimensional distances and angles were measured in both the original (absolute) models and the uniformly scaled models.

Results

With both measurement techniques, the CP in GHAI was more superior to that in CTRL (mean difference [MD] = 6.5 mm, P = .001 absolute; MD = 6 mm, P = .003 for scaled) and GHPI (MD = 2.4 mm, P = .015 absolute; MD = 2.2 mm, P = .021 scaled). The anterior coracoid coverage in GHAI was significantly lower than that in GHPI (MD = 3°, P = .009) and in CTRL (MD = 9°, P = .003). The lateral coracoid coverage in GHAI was lower than that in CTRL (MD = 13°, P = .008). The posterior acromial coverage of GHPI was decreased compared with that in both GHAI (MD = 5°, P = .006) and CTRL (MD = 5°, P = .018). All differences reported here remained statistically significant after Benjamini-Hochberg correction.

Conclusions

The position of the CP is more superior in patients with GHAI compared to GHPI and CTRL cohorts. Further research is needed to determine the biomechanical and clinical context in which a more superior CP position is associated with GHAI. Future studies should also determine whether, and under exactly which conditions, these morphologic findings should influence the surgical treatment decisions regarding recurrent anterior glenohumeral shoulder instability.

Incidence rates of primary dislocations of the glenohumeral joint are described between 21.9⁴² and 26.2²⁵ per 100,000 person-years. Recurrence of instability after primary shoulder dislocation is between 21% and 66.8%, and in a combination of male sex and young age, recurrence can be up to 80%.^21,32,38,50 The risk of developing severe glenohumeral osteoarthritis is between 10 and 20 times greater after glenohumeral dislocations than that in a population without dislocation.^7,9,28,30 This highlights the importance of adequate treatment of unstable shoulders. The potential role of morphologic variations for the stability of the shoulder and the etiology of antero-inferior glenohumeral instability (GHAI) is unclear. Predisposing anatomic factors for glenohumeral dislocation and its recurrence are still incompletely studied.^22,23

Owens et al³⁶ showed that an increased coracohumeral distance qualifies as an independent risk factor for GHAI. Jacxsens et al²³ showed that the coracoid process (CP) tip in relation to the center of the inferior glenoid best-fit circle is superior and more medial in traumatic GHAI cases than in healthy control shoulders. Rarely is there an association of GHAI with CP fractures.^31,37

Standard operative techniques such as the Latarjet^3,24,54 with complete osteotomy and transfer of the CP to the anterior glenoid rim or the Trillat^{10,13,44,46,47} procedure with incomplete osteotomy and re-orienting the CP tip inferiorly and medially may both prevent recurrence.^3,13,47,54 Together with the short head of the biceps brachii, the coracobrachialis creates the conjoined tendon which attaches to the CP.^2,14 The inferior transposition of the CP in the Latarjet and Trillat procedures, with the change in the conjoined tendon trajectory and the reconstruction of the anterior glenoid rim, improves the stability of the glenohumeral joint. These findings suggest that the CP with its attached soft-tissue structures may influence the stability of the glenohumeral joint.

It was the purpose of this paper to study whether the position and orientation of the CP could constitute a secondary restraint to anterior instability in analogy to the posterior acromion which has recently been shown to be a secondary restraint to posterior instability.^12,17, 18, 19 Therefore, the three-dimensional (3D) scapular morphology of the CP in patients with recurrent GHAI, in patients with recurrent posterior glenohumeral instability (GHPI), and in a healthy control (CTRL) cohort was analyzed. It was hypothesized that the CP tip position in GHAI would be more superior, more medial, and more anterior than that in a control group and in patients with GHPI, with the superior coracoid offset expected to be the most consistent and pronounced difference. The primary outcome measures were 3D angle and distance measurements of the CP tip relative to the glenoid center and glenoid plane in GHAI compared to GHPI and CTRL.

Materials and methods

Study design and ethical approval

This is a retrospective case-control study comparing shoulders with recurrent GHAI (cohort 1) to shoulders with recurrent GHPI (cohort 2) and shoulders without any clinical symptoms or pathologic radiographic findings (cohort 3). Approval was obtained from the responsible ethical committee (Kantonale Ethikkommission Kanton Bern, Project-ID 2024-02442). This study was designed and is reported in accordance with the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) recommendations for case-control studies.

Selection and description of cases

• •Cohort 1 (recurrent GHAI): we retrospectively reviewed all patients who had been treated non-operatively or operatively for recurrent GHAI between January 2019 and February 2025 in our institution. Inclusion criteria were typical diagnostic symptoms and/or diagnostic clinical findings, radiological documentation of recurrent GHAI (≥2 dislocation events), and availability of computed tomographic (CT) scans of the entire scapula. Diagnostic clinical findings were defined as ≥1 positive validated provocative instability test on physical examination. Among the included patients, a positive anterior apprehension test was documented in 55, a positive Jobe relocation test in 34, and a positive Gagey hyper-abduction test in 14. In 2 patients, clinical provocation testing was not feasible (eg, guarding/pain), and in 3 patients, no instability test was documented; in these 5 patients, the diagnosis was based on the documented history together with imaging findings. Radiographic documentation was preferably defined as a documented shoulder dislocation. Hill-Sachs lesions were also accepted. To exclude patients with anterior glenoid wear without anterior glenohumeral dislocations, anterior glenoid rim lesions were only accepted if they were accompanied by a Hill-Sachs lesion.⁵ Exclusion criteria were a lack of informed consent, primary or secondary glenohumeral osteoarthritis, previous shoulder infection, previous surgery on the affected shoulder, and rotator cuff pathologies. A total of 345 cases with GHAI were identified. In total, 187 cases were excluded because of missing informed consent, 42 due to unsatisfactory CT imaging, 11 due to prior surgery, 7 cases due to rotator cuff tendinopathy, 2 due to degenerative changes, and 19 cases for other reasons. Of these, 77 met all eligibility criteria. Of these 77 cases, 60 were matched by sex and age in a 2:1 ratio to the 30 identifiable and eligible cases with GHPI. In the anterior group (GHAI), 56 patients had surgery (46 Latarjet, 12 arthroscopic labral re-fixation, and 2 both).
• •Cohort 2 (recurrent GHPI): we retrospectively reviewed all patients who had been treated non-operatively and operatively for GHPI between January 2019 and February 2025 in our institution. Inclusion criteria were typical diagnostic symptoms and/or diagnostic clinical findings, radiological documentation of recurrent GHPI (≥2 subluxation events), and availability of a CT scan of the complete scapula. Diagnostic clinical findings were defined as ≥1 positive validated posterior provocative instability test on physical examination. Among the included patients, a positive Jerk test was documented in 20, a positive Scapinelli test in 17, a positive posterior drawer test in 9, and a positive posterior apprehension test in 4. Excluded were patients with no informed consent, primary or secondary osteoarthritis, previous shoulder infection, previous surgery on the affected shoulder, or rotator cuff pathologies. A total of 125 cases with GHPI were identified. In total, 60 cases were excluded because of missing informed consent, 23 due to unsatisfactory CT imaging, and 12 cases for other reasons. Of these, 30 met all eligibility criteria. These 30 cases were matched by age and sex to 60 shoulders of group 1 (GHAI). In the posterior group (GHPI), 17 had a surgery (9 arthroscopic, 4 posterior bone block [Scapinelli], and 4 posterior Bankart).
• •Cohort 3 (CTRL): the control cohort included 16 patients without any shoulder pathology. CT scans were taken for other reasons. These patients were the same cohort previously utilized for an analysis of the scapular morphology in certain patterns of glenohumeral osteoarthritis and full-thickness rotator cuff tears.³⁹ At the time of the examination, patients confirmed that they did not have previous trauma, dysfunction, or pain in the shoulder. The patients were followed up for two years after the CT scan to confirm the absence of symptoms. Three patients had to be excluded because they failed to provide informed consent.

All patient characteristics and demographic data can be found in Table I.

Table I.

Patient characteristics and demographic data.

Variable	GHAI (n = 60)	GHPI (n = 30)	CTRL (n = 13)	GHAI vs. GHPI P value	GHAI vs. CTRL P value	GHPI vs. CTRL P value
Age, yr, mean ± SD (95% CI)	28.5 ± 9.3 (26.1-30.9)	28.6 ± 9.3 (25.1-32.0)	61.7 ± 6.9 (57.5-65.9)	.986∗	<.001∗	<.001∗
Male, n (%)	44 (73%)	22 (73%)	9 (69%)	1.000†	.742†	1.000†
Right side, n (%)	32 (53%)	16 (53%)	3 (23%)	1.000†	.138†	.195†

GHAI, glenohumeral antero-inferior instability; GHPI, glenohumeral posterior instability; CTRL, control group; CI, confidence interval.

Revised in this version: P = .000 replaced with P < .001; all P values reported consistently to 3 decimal places.

^∗Non-parametric Mann-Whitney U test.

^†Fisher exact test. P values are reported to 3 decimals.

Scapula segmentation and preparation

All CT scans were exported as Digital Imaging and Communications in Medicine format (DICOM) files and were segmented using a deep learning–based algorithm.¹⁶ Each case's segmentation was manually corrected by one expert (S.P.S.), using ITK-SNAP (version 4.4.0; http://www.itksnap.org) and exported as a 3D mesh model. All mesh models were re-meshed and smoothed using 3D Slicer (version 5.11.0; nightly build 2025-11-10, http://slicer.org). Left-sided scapula models were mirrored to right-sided scapulae. To conduct measurements in a normalized scaled size, cases were isotropically scaled to the average geometric size.²⁹

Statistical shape modeling

Statistical shape modeling (SSM) is an established method for studying population-level shape variation and morphologic anatomic differences between subjects and cohorts.^15,23,39 SSM is a widely used method in the analysis of anatomic cohort data, providing the basis for comparing normal and pathologic 3D shape variations across groups.¹ For each case, corresponding mesh models were created using deformetrica.¹¹ Deformetrica is a groupwise correspondence method that is based on the large deformation diffeomorphic metric mapping framework to create corresponding mesh models for each case.^11,15,33 One random shape from the CTRL group was chosen to be the reference shape. This shape was smoothed to remove noise and re-meshed to approximately 30,000 vertices. According to Lüthi et al,²⁶ in the absence of prior knowledge, it was assumed that the chosen reference shape is close to the hypothetical average shape of the modeled class of shapes. Scalismo (version 0.92; https://scalismo.org) was used to perform SSM and principal component analysis and to visualize the deformable models.²⁷ For each group, surface mean shapes were created to qualitatively compare the differences between the 3 cohorts. The resulting angular and distance values were therefore compared with previously published normal angle and distance measurements of other CTRL mean shapes.^4,23

Landmark detection and scapular alignment

For a quantitative comparison of distance and angle measurements, standard anatomic landmarks were generated on every segment using a second deep learning network for morphologic analysis.¹⁶ To quantify differences between subjects, all cases had to be aligned specifically. The superior-inferior axis was defined through the supraglenoid tubercle landmark and the inferior glenoid rim landmark. The center point between the supraglenoid tubercle and the inferior glenoid rim landmark was projected orthogonally to the superior-inferior axis and to the vector between the anterior and posterior glenoid rim landmarks onto the glenoid surface and was then defined as the glenoid center and set as the origin of the coordinate system. The superior-inferior z-axis was defined through the origin and parallel to the superior-inferior axis. The glenoid plane was defined as the plane containing the origin, parallel to the z-axis and to the vector between the anterior and posterior glenoid rim landmarks. The anterior-posterior y-axis was defined in the glenoid plane and perpendicular to the z-axis. The lateral-medial x-axis was defined perpendicular to the glenoid plane. All cases were then aligned to the glenoid center and its plane (Fig. 1).

Figure 1.

(A) Visualization of all placed anatomic landmarks. Three landmarks were placed on anterior, inferior, and posterior glenoid rim (C1, C2, and C3). The glenoid Center was created as described in the Landmark detection and scapular alignment section and served as the origin (O). Other landmarks were placed on the SGT, the LCT, and the acromial PLC. (B) The PAC (blue) was measured between the xz-plane and a plane created through the origin, parallel to the x-axis and through the PLC. The ACC (green) was measured between the xz-plane and a plane created through the origin, parallel to the x-axis and tangential to the coracoid process. (C) The LCC (red) was measured between the yz-plane and a plane created through the origin, parallel to the y-axis and tangential to the coracoid process. PLC, posterior lateral corner; SGT, superior glenoid tubercle; PAC, posterior acromion coverage angle; ACC, anterior coracoid coverage angle; LCC, lateral coracoid coverage; LCT, lateral coracoid tip.

Morphometric measurements

Distance and angle measurements were performed on all absolute and scaled models. The coracoid apex offset in relation to the glenoid center and its plane was measured. In all cases, the anterior coracoid coverage and posterior acromion coverage were measured in the sagittal plane (yz-plane), and the lateral coracoid coverage was measured in the coronal plane (xz-plane) (Fig. 1).

Statistical analysis

The sample size was determined a priori for the primary outcome, the superior offset of the coracoid tip in the GHAI vs. the CTRL cohort. According to Jacxsens et al,²³ a mean difference (MD) in coracoid superior offset of 7 mm and a common standard deviation (SD) of 5 mm between the GHAI cohort and the CTRL cohort were assumed (Cohen's d = 1.4). For a two-tailed significance level (alpha) of 0.05 and a target power of 0.95, by applying the planned 4:2:1 enrollment ratio across the GHAI, GHPI, and CTRL cohorts, an independent two-sample t-test required a minimum of n = 40 (group 1, GHAI), n = 20 (group 2, GHPI), and n = 10 (group 3, CTRL) shoulders. The realized sample (60 GHAI, 30 GHPI, and 13 CTRL) therefore exceeded the planned minimum, providing a power of approximately 0.99 for the primary comparison. Data normality was tested using Shapiro-Wilk test. Values are presented in means, SD, and 95% confidence intervals. P values below .05 were considered statistically significant. Using 2-way analysis of variance, differences between the cohorts were assessed. For pairwise comparisons between the 3 groups (GHAI vs. GHPI; GHAI vs. CTRL; GHPI vs. CTRL), Welch's unequal variance t-test⁴⁰ was used for continuous normally distributed data, the non-parametric Mann-Whitney U test was used for continuous non-normally distributed data, and Fisher's exact test was used for categorical data. Because multiple pairwise comparisons were performed across the 3 cohorts, P values for the pairwise morphometric comparisons were additionally adjusted for multiple tests using the Benjamini-Hochberg false discovery rate procedure; both unadjusted and adjusted P values are reported, and an adjusted P < .05 was considered significant. The superior coracoid offset in GHAI vs. CTRL was pre-specified as the primary outcome.

Results

At the time of examination, the GHAI and GHPI cohorts were both, with an average age of 33 years, younger than the CTRL cohort (P < .001). There were no significant differences in sex or sidedness between the cohorts (sex, P ≥ .742; sidedness, P ≥ .138). All demographic patient data can be found in Table I.

Comparing standard measurements of the CTRL mean shape model with measurements published by Beeler et al⁴ and Jacxsens et al,²³ the latter were always within the respective SDs.

GHAI vs. GHPI

Qualitatively comparing the mean shape of GHAI with that of GHPI, there was a relatively large difference in shape around the CP, acromion, and glenoid rim (Fig. 2). A quantitative comparison of coracoid offset measurements between GHAI and GHPI showed that the CP tip was on average 2.4 mm more superior (SD GHAI = 4.6 mm; SD GHPI = 4.2 mm; P = .015, adjusted P = .034), 2.4 mm less lateral (SD GHAI = 4.5 mm; SD GHPI = 4.3 mm; P = .014, adjusted P = .034), and 1.5 mm less anterior (SD GHAI = 3.7 mm; SD GHPI = 3 mm; P = .043, adjusted P = .073) in absolute measurements and on average 2.2 mm more superior (SD GHAI = 4.2 mm; SD GHPI = 4.3 mm; P = .021, adjusted P = .041) and 2.5 mm less lateral (SD GHAI = 4.6 mm; SD GHPI = 4.3 mm; P = .013, adjusted P = .034) in scaled measurements. In GHAI, the anterior coracoid coverage angle was smaller than that in GHPI (MD = 3°; SD GHAI = 6°; SD GHPI = 6°; P = .009, adjusted P = .027). The posterior acromial coverage angle in GHAI was larger than that in GHPI (MD = 5°; SD GHAI = 6°; SD GHPI = 7°; P = .006, adjusted P = .023).

Figure 2.

The statistical shape models were qualitatively assessed pairwise to identify shape differences between cohorts and to guide distance and angle measurements. The light blue scapula represents the mean shape of the GHAI cohort, the magenta scapula represents the mean shape of the GHPI cohort, and the yellow scapula represents the mean shape of the CTRL cohort. Qualitative morphologic differences are described in Results. GHAI, antero-inferior glenohumeral instability; CTRL, control; GHPI, posterior glenohumeral instability.

GHAI vs. CTRL

Qualitatively comparing the mean shape of GHAI with that of CTRL, there was a relatively large difference in shape around the CP, with a smaller difference around the anterior glenoid rim (Fig. 2). A quantitative comparison of coracoid offset measurements between GHAI and CTRL revealed the CP tip to be, on average, 6.5 mm more superior (SD GHAI = 4.6 mm; SD CTRL = 5.4 mm; P = .001, adjusted P = .014) in absolute measurements and, on average, 6 mm more superior (SD GHAI = 4.2 mm; SD CTRL = 5.8 mm; P = .003, adjusted P = .016) and 1.2 mm less lateral (SD GHAI = 4.6 mm; SD CTRL = 4.2 mm; P = .049, adjusted P = .073) in scaled measurements. In GHAI, the lateral coracoid coverage angle was smaller than that in CTRL (MD = 13°; SD GHAI = 10°; SD CTRL = 15°; P = .008, adjusted P = .027). The anterior coracoid coverage angle in GHAI was smaller than that in CTRL (MD = 9°; SD GHAI = 6°; SD CTRL = 9°; P = .003, adjusted P = .016).

GHPI vs. CTRL

Comparing coracoid offset measurements of GHPI with those of CTRL, the CP tip was on average 5.1 mm less lateral (SD GHPI = 4.3 mm; SD CTRL = 4.8 mm; P = .003, adjusted P = .016) and 4.1 mm more superior (SD GHPI = 4.2 mm; SD CTRL = 5.4 mm; P = .027, adjusted P = .049) in absolute measurements and 5.2 mm less lateral (SD GHPI = 4.3 mm; SD CTRL = 4.2 mm; P = .001, adjusted P = .014) in scaled measurements. In GHPI, the lateral coracoid coverage angle was smaller than that in CTRL (MD = 15°; SD GHPI = 10°; SD CTRL = 15°; P = .005, adjusted P = .023). The posterior acromial coverage angle in GHPI was also smaller than that in CTRL (MD = 5°; SD GHPI = 7°; SD CTRL = 5°; P = .018, adjusted P = .037).

Detailed results are found in Tables II and III. Qualitative mean shape comparisons are illustrated in Figure 2. Superior CP offset measurement, posterior acromion coverage angle, and anterior coracoid coverage angle are illustrated in Figure 3. After Benjamini-Hochberg adjustment for multiple comparisons, the superior coracoid offset differences and the principal coverage-angle differences remained significant, whereas the small anterior coracoid offset difference between GHAI and GHPI and the scaled lateral offset difference between GHAI and CTRL did not retain significance (Tables II and III).

Table II.

Distance measurements (coracoid apex offset).

Measurement	GHAI (n = 60) Mean ± SD (95% CI)	GHPI (n = 30) Mean ± SD (95% CI)	CTRL (n = 13) Mean ± SD (95% CI)	GHAI vs. GHPI P (adj. P) values∗	GHAI vs. CTRL P (adj. P) values∗	GHPI vs. CTRL P (adj. P) values∗
Offset measurements, scaled (mm)
Lateral	13.8 ± 4.6 (12.6-15.0)	11.3 ± 4.3 (9.7-12.9)	16.5 ± 4.2 (13.9-19.0)	.013 (.034)	.057 (.073)	.001 (.014)
Anterior	28.1 ± 3.0 (27.3-28.8)	29.3 ± 2.7 (28.3-30.3)	29.3 ± 1.6 (28.3-30.3)	.052 (.073)	.049 (.073)	.993 (.993)
Superior	24.3 ± 4.2 (23.2-25.4)	22.1 ± 4.3 (20.5-23.6)	18.3 ± 5.8 (14.9-21.8)	.021 (.041)	.003 (.016)	.051 (.073)
Offset measurements, absolute (mm)
Lateral	13.7 ± 4.5 (12.5-14.9)	11.3 ± 4.3 (9.7-12.8)	16.4 ± 4.8 (13.5-19.3)	.014 (.034)	.081 (.099)	.003 (.016)
Anterior	28.1 ± 3.7 (27.1-29.0)	29.6 ± 3.0 (28.4-30.7)	28.9 ± 2.9 (27.1-30.6)	.043 (.073)	.395 (.464)	.488 (.549)
Superior	24.3 ± 4.6 (23.1-25.5)	21.9 ± 4.2 (20.3-23.5)	17.8 ± 5.4 (14.6-21.1)	.015 (.034)	.001 (.014)	.027 (.049)

GHAI, glenohumeral antero-inferior instability; GHPI, glenohumeral posterior instability; CTRL, control group; CI, confidence interval; CT, computed tomography.

Revised in this version: Benjamini-Hochberg adjusted P (q) values added in parentheses; P values reported consistently to 3 decimals. Superior GHAI vs. CTRL, offset was the pre-specified primary outcome.

^∗All pairwise comparisons: Welch unequal-variance t-test. Distances reported to 0.1 mm (CT-based measurement resolution).

Table III.

Angle measurements (coracoacromial coverage).

Measurement	GHAI (n = 60) Mean ± SD (95% CI)	GHPI (n = 30) Mean ± SD (95% CI)	CTRL (n = 13) Mean ± SD (95% CI)	GHAI vs. GHPI P (adj. P) values	GHAI vs. CTRL P (adj. P) values	GHPI vs. CTRL P (adj. P) values
Lateral coracoid coverage	33 ± 10 (30-35)	31 ± 10 (28-35)	46 ± 15 (37-54)	.548 (.592)∗	.008 (.027)∗	.005 (.023)∗
Anterior coracoid coverage	53 ± 6 (51-54)	56 ± 6 (54-59)	62 ± 9 (57-67)	.009 (.027)∗	.003 (.016)∗	.055 (.073)∗
Posterior acromial coverage	80 ± 6 (78-81)	75 ± 7 (73-78)	80 ± 5 (77-83)	.006 (.023)†	.779 (.809)†	.018 (.037)∗

GHAI, glenohumeral antero-inferior instability; GHPI, glenohumeral posterior instability; CTRL, control group; CI, confidence interval.

Revised in this version: angle values rounded to whole degrees (sub-degree precision is not measurable); Benjamini-Hochberg adjusted P (q) values added in parentheses; P values reported consistently to 3 decimals.

^∗Welch unequal-variance t-test.

^†Non-parametric Mann-Whitney U test. Angles in degrees.

Figure 3.

Sagittal view of the 3 mean mesh models showing the most important results. The blue scapula represents the mean shape of the GHAI cohort, the magenta scapula represents the mean shape of the GHPI cohort, and the yellow scapula represents the mean shape of the CTRL cohort. Illustrated are the PAC in blue, the ACC in green, and the superior coracoid process offset (height relative to the glenoid center and its plane) in light gray. In GHAI, the superior coracoid process offset is increased, and the ACC is decreased compared to GHPI and CTRL. Also adding to the morphologic analysis, the PAC in GHPI is decreased in comparison to that in GHAI and CTRL. PLC, posterior lateral corner; SGT, superior glenoid tubercle; PAC, posterior acromion coverage angle; ACC, anterior coracoid coverage angle; LCT, lateral coracoid tip; GHAI, antero-inferior glenohumeral instability; GHPI, posterior glenohumeral instability; CTRL, control; SUP, superior.

Discussion

This study compares the 3D CP position of a GHAI cohort with that of a GHPI cohort and stable shoulders to identify morphologic conditions that are associated with different directions of instability. Previously, coracoid morphology in posterior instability has been investigated by Testa et al,⁴³ who reported significant differences in the coronal coracoid angle, coracoacromial length, and the inferior coracoid coverage between posterior stabilization patients, anterior instability patients, and a comparison cohort using magnetic resonance imaging–based measurements. However, their methods relied on 2D slice-based measurements. This study complements this work by quantifying coracoid morphology in 3 dimensions on SSMs from complete-scapula CT, using a scapular reference frame independent of patient positioning. By comparing GHAI and GHPI with CTRL, the aim was to identify pathologically associated morphology that is not present in normal, stable shoulders.

The hypothesis of a more superior, more medial, and more anterior CP in GHAI was partially confirmed. The tip of the CP was in both qualitative SSM-based and quantitative 3D measurements more superior in GHAI than in GHPI and CTRL (Fig. 3). The anterior coracoid coverage in GHAI was significantly decreased compared with that in GHPI and CTRL. In addition, the lateral coracoid coverage in GHAI was significantly decreased compared only to that in CTRL. In comparison with GHPI and CTRL, it could not be confirmed that a more anterior and medial position of the CP is associated with GHAI. Because of the retrospective case-control design, these findings demonstrate an anatomic association and cannot establish whether the coracoid position is an independent predisposing factor for instability or a secondary morphologic characteristic; they should therefore be confirmed in prospective longitudinal studies.

In all offset measurements, the CP tip in the GHPI was more medial than that in both the GHAI and CTRL cohorts. The glenoid version of the mean GHPI SSM was approximately −8°, compared with −3° for the mean GHAI SSM. Consistent with this, previous studies have shown that GHPI is associated with increased glenoid retroversion.^4,35,48,51 The reference coordinate system for this study was aligned to the glenoid; therefore, the more medial position of the CP tip observed in the GHPI cohort may result from the glenoid alignment method applied in this study. In contradiction to the results of this study, the recently published results by Testa et al⁴³ showed that patients with GHPI had a more inferior CP position than the GHAI and CTRL cohorts. However, direct comparability between the 2 studies is limited because of the use of substantially different measurement techniques.

The differences in superior CP position between GHAI and CTRL are consistent with values reported by Jacxsens et al²³ and with the findings of Owens et al.³⁶ However, coracohumeral distance measurements depend on humeral head position and therefore cannot be directly compared with the CP offset measurements in this study.

The association of CP fractures with antero-inferior dislocations suggests that in shoulders with a more inferior position of the CP, the CP could act as a secondary anterior restraint for the humeral head during anterior translation on the glenoid surface.^31,37

The findings of a more superior CP in GHAI may be associated with the higher recurrence rates after soft-tissue procedures (eg, arthroscopic Bankart repair) compared with the rates following Latarjet or Trillat procedures, in which the anatomy of the CP is permanently changed.^3,54 However, in the Latarjet procedure, the CP is used to augment the anterior glenoid rim, which removes its function as a secondary restraint but enlarges the glenoid surface.²⁴ A secondary effect is a change in the trajectory of the conjoined tendon, which would be comparable to the lower trajectory in more inferior CP positions found more frequently in GHPI and CTRL cohorts.⁵²

The success of the Trillat procedure in preventing recurrent GHAI is in agreement with the findings in this study.^{6,10,13,20,41,45,47,49} In the Trillat procedure, simple lowering of the CP has been shown to be effective in preventing recurrent anterior shoulder dislocations. The effectiveness of this procedure in elderly patients with GHAI associated with massive rotator cuff tears is easily conceivable. Superior migration of the humeral head might result in a relatively lower position of the CP tip, thereby increasing the effectiveness of the CP as a secondary anterior restraint.^6,41

Important risk factors for GHAI are young age, male sex, high activity level, ligamentous laxity, and tall and thin glenoids.^8,34,36,53 Particularly in cases of recurrent GHAI, established risk factors include glenoid or humeral head bone loss.³² In the context of these risk factors, a more superior position of the CP, if it is already present before the initial dislocation, which the present cross-sectional design cannot establish, could be associated with both the initial antero-inferior dislocation and the recurrent GHAI. These findings contribute to a better understanding of the potential etiology and pathophysiology of GHAI and may inform future applications such as 3D-planned coracoid osteotomy, ultimately improving treatments for patients.

At present, these findings are hypothesis-generating and do not yet justify routine measurement of the coracoid position to guide the choice between non-operative and operative treatment after a first-time dislocation. They do, however, identify the position of the coracoid as a potentially modifiable osseous factor that should be examined prospectively for its association with recurrence and its value in surgical decision-making.

Additionally and in line with previous studies, the identified decrease in posterior acromial coverage was specific to GHPI.^12,17, 18, 19

The study has limitations. The validity of the generated CTRL mean shape model may be limited by the initial number of shape samples, as all available samples from a previously validated fixed cohort were used. This validity of the control group is supported by the similar morphologic values found in similar studies by Beeler et al⁴ and Jacxsens et al.²³ A statistical power analysis showed that both comparisons of superior CP offset measurements between GHAI and CTRL had a power greater than 95%. Another limitation was that the CTRL cohort was 33 years older (P < .001) than the other 2 cohorts; however, degenerative changes of the glenohumeral joint were excluded. Matching the CTRL cohort to the other 2 cohorts was not possible. Furthermore, the control cohort was drawn from a separate, previously published cohort and could not be matched to the instability groups for age, sex, or activity level; although scapular osseous morphology is generally considered to remain essentially stable throughout adulthood, this age and provenance mismatch remain a source of potential bias, and comparisons involving the control cohort should be interpreted with caution.

The MDs of superior CP offset between GHAI and GHPI were only about 2.4 mm, which is about 6.4% of the mean glenoid height. The MD in the superior CP offset of GHAI and normal shoulders is about 6.5 mm, which is already 17.7% of the mean glenoid height. Although these MDs alone may not be clinically relevant, substantial variability in CP height was observed. Individuals in the GHAI cohort showed superior coracoid offset up to 17.4 mm above the mean CTRL CP height, corresponding to 47.5% of the glenoid height, which may well be clinically relevant (Fig. 4). It must be emphasized that the mean between-group differences, particularly the 2.4 mm superior offset between GHAI and GHPI, are small in absolute terms, and whether differences of this magnitude have any direct biomechanical or clinical consequence remains speculative and requires dedicated biomechanical and prospective clinical confirmation.

Figure 4.

Case with the greatest coracoid process superior offset of the antero-inferior glenohumeral instability cohort. (A) Sagittal view of the subject scapula. The superior coracoid process offset is measured to be 35.2 mm in this case. A large bony Bankart lesion is also present. (B) Side view of the subject scapula. In summary, this example raises the question of whether the superior coracoid process offset, the anterior glenoid rim defect, or their combination is more strongly associated with antero-inferior glenohumeral instability. SGT, superior glenoid tubercle; LCT, lateral coracoid tip; SUP, superior.

A further limitation of this study was that the exact number of dislocation events per patient could not be reliably ascertained in all cases; we can confirm only that every included patient had sustained at least 2 episodes, consistent with the inclusion criterion.

The extent of glenoid and humeral head bone loss was not quantified, and patients were not stratified according to bipolar bone loss. Because such osseous lesions are a consequence of the dislocation rather than its cause, they were not the focus of the present study, which addressed morphology predisposing to the initial event; nevertheless, they are major modifiers of recurrence and warrant dedicated investigation. Because glenoid bone loss can additionally alter the glenoid landmarks that define the reference coordinate system, unquantified bone loss may also have affected the apparent position of the coracoid relative to the glenoid, an effect that could not be excluded in the present study. Patients with clinically multi-directional instability were excluded by design although later development of bi- or multi-directional instability could not be excluded at the time of assessment. Finally, a substantial proportion of otherwise eligible patients were excluded because informed consent was not available; although their morphology could not be characterized, this may introduce a selection bias, and because a large proportion of screened patients, particularly those without available informed consent or without complete CT imaging, were excluded, the analyzed sample may not be fully representative of the source instability population, and the direction of any resulting bias cannot be determined. In addition, potential confounders such as arm dominance, activity level, sport participation, and generalized ligamentous hyper-laxity were not systematically recorded in this retrospective cohort and could therefore not be entered into a multi-variable model; the analyses were consequently limited to unadjusted pairwise comparisons, and a prospective study with multi-variable adjustment will be required to determine whether the observed association between coracoid position and instability is independent of these factors.

Conclusion

The position of the CP in patients with GHAI is more superior to that in GHPI and CTRL cohorts. Further research is needed to determine the biomechanical and clinical context in which a more superior CP position is associated with antero-inferior glenohumeral instability. Future studies should also determine whether, and under exactly which conditions, these morphologic findings should influence the surgical treatment decisions for recurrent anterior glenohumeral shoulder instability.

Disclaimers

Funding: Matthias A. Zumstein declares research funding from Medacta and also form Chemex, unrelated to the submitted work. However, no funding was received for the conduct of this study.

Conflicts of interest: Matthias A. Zumstein declares grants from Medacta. Furthermore, he is a designer for VIMS, Zimmer, and Bonebridge and receives non-financial support from Angiocrine Biosciences, unrelated to the submitted work. Other than here declared, the author, his immediate family, and any research foundation with which he is affiliated have not received any further financial payments or other benefits from any commercial entity related to the subject of this article. Any additional authors, their immediate families, and any research foundation with which they are affiliated have not received any financial payments or other benefits from any commercial entity related to the subject of this article.